Biometric systems use a person's distinguishing biological and/or behavioral traits to recognize and/or identify a user. Voice recognition, also known as voiceprint recognition, is a performance biometric, i.e., a user performs a task, speaking, and the biometric system uses the user's speech to verify the user, or to identify the user. Using a user's voiceprint for verification and/or identification can obviate the need for a user to remember or maintain knowledge-based and/or possession-based access-control elements such as, for example, a specific password or token generator, respectively, since the user's voiceprint is used by the biometric system.
Biometric systems can be designed to recognize signal interference, noise variations, mimicry attempts, and/or ordinary voice changes, for example, voice changes that arise due to time of day, illness, fatigue, and the like. Therefore, a well-designed biometric system can provide reliable access-control for some applications.
In practice, biometric systems used to verify users can require the user to complete a registration step, wherein the user's voiceprint is generated by the biometric system and stored for future use. When a user attempts to authenticate with the biometric system, the biometric system can generate a new voiceprint from the user attempting to authenticate, and compare the new voiceprint with the stored voiceprint to determine if there is a match. Similarly, use of biometric systems to identify users can require a collection of voiceprints against which a new voiceprint can be compared. By searching for and identifying a match, the biometric system can determine the identity of the user whose voiceprint is being analyzed for a match.
Voiceprint recognition systems can use acoustic patterns of speech to generate a voiceprint. The acoustic patterns can include anatomical and behavioral characteristics. With respect to anatomical characteristics, the vocal tract is typically described as the speech organs that exist beyond the vocal chords, also known as the vocal folds. The speech organs are generally described as including the laryngeal pharynx, the oral pharynx, the oral cavity, i.e., the void bound by the velum, palate, tongue, and lips, the nasal pharynx, and the nasal cavity, i.e., the void extending from the nasal pharynx to the nostrils. The speech organs can affect the frequency content of speech by altering resonances of the frequency content. As such, the shape and size of the speech organs can often be estimated by analyzing the frequency content of a voice.
Behavioral characteristics, such as voice pitch, speech rate, inflection, dialect, speaking style, accent, and the like, can also be recognized by a biometric system. Such behavioral characteristics are sometimes referred to as “behavioral biometrics” because these characteristics are based on learned speech patterns more than on anatomical structure of the user. The anatomical or behavioral characteristics, and/or a combination thereof, can be identified and used to generate a user-specific voiceprint.
Biometric systems may need to run extremely complex mathematical calculations on a voice signal to determine these and other characteristics, and to compare an obtained voiceprint to a voiceprint stored as a reference. Additional calculations may be undertaken to correct for, or to identify, interference, mimicry, signal degradation, channel shifting, and the like. Enormous amounts of data and data analysis must often be analyzed and/or undertaken to successfully implement a biometric system as a verification or identification access-control mechanism.